背景:长链非编码RNA(lncRNA)是一类涉及重要生物学过程的分子。已经提供了广泛的努力,以在lncRNA水平上更深入地了解疾病机制,指导检测用于疾病诊断的生物标志物,治疗,预后和预防。不幸的是,由于成本和时间的复杂性,通过传统生物学实验验证的可能的疾病相关lncRNAs数量非常有限.预测疾病-lncRNA关联的计算方法允许识别最有希望在实验室中验证的候选物。降低成本和耗时。
结果:我们提出了预测lncRNA-疾病关联的新方法,所有人都分享了探索lncRNAs之间关联的想法,其他中间分子(例如,miRNA)和疾病,适当地由三部图表示。的确,虽然只有少数lncRNA与疾病的关联仍然是已知的,lncRNAs和其他分子之间的大量相互作用,以及后者与疾病的联系,是可用的。这里介绍的第一种方法,NGH,依赖于对三部图进行的邻域分析,建立在lncRNAs上,miRNA和疾病。第二种方法(CF)依赖于协同过滤;第三种方法(NGH-CF)通过协同过滤增强NGH来获得。所提出的方法已经在合成数据和实际数据上得到了验证,并与文献中的其他方法进行了比较。它的结果是邻域分析允许超越竞争对手,当与协同过滤相结合时,预测精度进一步提高,评分AUC值等于0966。
背景:源代码和示例数据集可从以下网址获得:https://github.com/marybonomo/LDAsPredictionApproaches。git.
BACKGROUND: Long non-coding RNAs (lncRNAs) are a class of molecules involved in important biological processes. Extensive efforts have been provided to get deeper understanding of disease mechanisms at the lncRNA level, guiding towards the detection of biomarkers for disease diagnosis, treatment, prognosis and prevention. Unfortunately, due to costs and time complexity, the number of possible disease-related lncRNAs verified by traditional biological experiments is very limited. Computational approaches for the prediction of disease-lncRNA associations allow to identify the most promising candidates to be verified in laboratory, reducing costs and time consuming.
RESULTS: We propose novel approaches for the prediction of lncRNA-disease associations, all sharing the idea of exploring associations among lncRNAs, other intermediate molecules (e.g., miRNAs) and diseases, suitably represented by tripartite graphs. Indeed, while only a few lncRNA-disease associations are still known, plenty of interactions between lncRNAs and other molecules, as well as associations of the latters with diseases, are available. A first approach presented here, NGH, relies on neighborhood analysis performed on a tripartite graph, built upon lncRNAs, miRNAs and diseases. A second approach (CF) relies on collaborative filtering; a third approach (NGH-CF) is obtained boosting NGH by collaborative filtering. The proposed approaches have been validated on both synthetic and real data, and compared against other methods from the literature. It results that neighborhood analysis allows to outperform competitors, and when it is combined with collaborative filtering the prediction accuracy further improves, scoring a value of AUC equal to 0966.
BACKGROUND: Source code and sample datasets are available at: https://github.com/marybonomo/LDAsPredictionApproaches.git.